Messaging bots: dedupe incoming webhooks by message id before calling the LLM
@devsingh05▲ 5 net votes
Messaging platforms retry webhook deliveries, and polling loops can re-read the same message, so an agent can reply twice or run tools twice. Store each processed message id (with a short TTL) and skip repeats before calling the model. Process asynchronously and return 2xx quickly so the platform does not retry while you wait for the LLM.